THE EFFECTS of WORLD WAR II on ECONOMIC and HEALTH OUTCOMES ACROSS EUROPE Iris Kesternich, Bettina Siﬂinger, James P

THE EFFECTS OF WORLD WAR II ON ECONOMIC AND HEALTH OUTCOMES ACROSS EUROPE Iris Kesternich, Bettina Siﬂinger, James P. Smith, and Joachim K. Winter*

Abstract—We investigate long-run effects of World War II on socioeco- able. SHARE covers representative samples of the popula- nomic status and health of older individuals in Europe. We analyze data from SHARELIFE, a retrospective survey conducted as part of SHARE tion aged 50 and over in thirteen European countries, with in Europe in 2009. SHARELIFE provides detailed data on events in child- about 20,000 observations. We also collected external data hood during and after the war for over 20,000 individuals in thirteen on casualties, timing and location of combat action, yearly European countries. We construct several measures of war exposure: experience of dispossession, persecution, combat in local areas, and hun- GDP by country, population movements, and male-female ger periods. Exposure to war and, more important, to individual-level population ratios. To our individual-level analysis of the shocks caused by the war significantly predicts economic and health out- multidimensional effects of a major shock that affected life comes at older ages. circumstances, we add new dimensions to a rapidly increasing literature that aims to explain the causes of health and wealth gradients in labor and health economics (see Deaton, I. Introduction 2007; Smith, 2009a; Heckman, 2012). SHARE not only measures major contemporaneous eco- ORLD War II (WWII) was one of the major trans- nomic and health outcomes of adults over age 50 in these W formative events of the twentieth century, with 39 European countries but includes retrospective modules million deaths in Europe alone. Large amounts of physical meant to capture salient parts of early life experiences, capital were destroyed through six years of ground battles including those related to the war. There are no microeco- and bombing. Many individuals were forced to abandon or nomic panel data in either the United States or in Europe give up their property without compensation and move to that have prospectively tracked people for that long a time new lands. Periods of hunger became more common even period.1 The coexistence of current prospective data com- in relatively prosperous Western Europe. Families were bined with retrospective data on key events that preceded separated for long periods of time, and many children lost the survey baseline opens up important new research oppor- their fathers. Many, including young children, personally tunities not before possible, and not simply those associated witnessed the horrors of war as battles and bombing took with the war. Since the end of World War II, Western conti- place in the very areas where they lived. Horrendous crimes nental Europe has had a rich and sometime tumultuous eco- against humanity were committed. Due to the war, political nomic and political history, the effects of which on its resi- and economic systems in many countries were permanently dents are not well documented. altered. There is legitimate concern about the quality of recall In this paper, we investigate the long-run effects of data, particularly for time periods decades in the past. But World War II on late-life economic and health outcomes in that concern has been lessened by a realization that recall of Western continental Europe (health, education, labor mar- events during childhood is better than for other periods of ket outcomes, and marriage). We explore several channels life, particularly if events are as salient, they certainly are in through which this war might have influenced individual this application. Smith (2009b) investigated several quality lives and document which groups of the population were markers and showed that his childhood health instrument most affected. Our research relies on retrospective life data was successful in matching known secular trends in child- from the European Survey of Health, Aging, and Retire- hood illnesses decades in the past.2 Moreover, we provide ment in Europe (SHARE) that have recently become avail- evidence in this paper that these recalled events in the SHARE retrospective about the war match the historical Received for publication December 17, 2011. Revision accepted for record. publication August 30, 2012. * Kesternich and Siflinger: University of Munich; Smith: RAND; Win- One aim of the paper is to illustrate how such retrospec- ter: University of Munich. tive life data can further our understanding of effects of We thank Alexander Danzer, Angus Deaton, Benjamin Friedman, Edward Glaeser, Dirk Jenter, Olmo Silva, Till von Wacher, participants of early-life conditions as affected by large external shocks the conference on Fetal Origins, Early Childhood Exposure, and Famine at such as war. The literature measuring impacts of macro- Princeton University, September 2011, as well as seminar audiences at events mostly has used ‘‘natural experiments’’ such as wars Boston University, Harvard University, and the University of Cologne for helpful comments on earlier versions. Sarah Lehner and Johanna Sophie Quis provided excellent research assistance. Kesternich acknowledges financial support from the DFG through SFB/TR 15. Smith is supported by 1 PSID, the longest microeconomic panel, began in 1968 more than various grants from NIA. This paper uses data from SHARELIFE release twenty years after World War II. The longest-running European microe- 1, as of November 24, 2010, or SHARE release 2.5.0, as of May 24, 2011. conomic panel, GSEOP, began in 1984, almost forty years after the war. The SHARE data collection has been primarily funded by the European 2 There was also no evidence of backward attribution of new episodes Commission through the fifth framework and sixth framework program. of adult health problems into a revaluation of childhood health. Adult SHARELIFE was supported through the seventh framework program. respondents whose health deteriorated between PSID waves were no Additional funding from the U.S. National Institute on Aging, as well as more likely than before to say their childhood health was not good or to from various national sources, is gratefully acknowledged. cite additional childhood health problems (Smith, 2009b)

The Review of Economics and Statistics, March 2014, 96(1): 103–118 Ó 2014 by the President and Fellows of Harvard College and the Massachusetts Institute of Technology

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/REST_a_00353 by guest on 29 September 2021 104 THE REVIEW OF ECONOMICS AND STATISTICS

or famines to study effects of early-life conditions at the In addition to a standard set of demographic attributes aggregate level. Largely due to data reasons beyond their (age, marital status, education), SHARE data include health control, the studies of which we are aware could not use variables (self-reported health, health conditions, health individual-level measures of whether a particular person behaviors), psychological variables (such as depression and was affected by the war and through which channel. Retro- well-being), economic variables (current work activity, spective life data, such as those from SHARE, contain sources and composition of current income), and net wealth detailed information and provide the opportunity to study (including housing, cars, and all financial assets (stocks, that issue. bonds), and cash minus all debts). Analyzing different outcomes is a first step in understand- SHARE’s third wave of data collection, SHARELIFE, ing the channels and mechanisms by which wars affect peo- collected detailed retrospective life histories in thirteen ple’s lives. Another possibility is using different measures of countries (Poland and the Czech Republic were added in war exposure such as the closeness of combat. We construct wave 2) in 2008–2009. SHARELIFE was based on life his- such measures from external data sources. In addition, tory calendar (LHC) methods. The interview starts with the SHARE data contain retrospective questions on several pos- names and birth dates of the respondent’s children (and sible channels of war exposure: hunger, the absence of the other information about them, including any deaths), fol- father, dispossession, and persecution. lowed by a full partner and residential history. This infor- Given the scale of the war and number of ways it funda- mation is used to aid in dating all other events. mentally changed the world, the existing economic literature The information in the life history includes family com- using World War II as a natural experiment is surprisingly position and type of home (number of rooms, running thin. Moreover, the literature that does exist is relatively water, toilet, and so on), number of books, and occupation recent and more American in context than European. This of father. These measures were used to create an index of may reflect the fact that the popularity of the natural experi- childhood SES at age 10. A childhood health history is also ments framework in economics itself postdated the war by included based on the Smith module included in the PSID many decades. Still, it does suggest that excellent research and HRS that queries about individual specific childhood opportunities remain, especially given the wide diversity of diseases and an overall subjective evaluation of childhood European experiences in World War II. health status (Smith, 2009b). In addition, respondents are This paper is divided into six sections. The next highlights asked about childhood immunizations and hunger during the main attributes of SHARE data and the additional data childhood. Adult health histories and job and income his- we collected for this research. Section III sets the stage for tories were also collected. Moreover, SHARELIFE pro- our analysis by presenting evidence of possible changes on vides detailed data on within-country region of residence which long-term effects of World War II may operate. The and housing during the full life of the respondents (child- fourth section summarizes statistical models that capture hood and adulthood). impacts of the experience of World War on the individual adult labor market, demographic, and health outcomes. We B. Other Data Sources also present our models of the influence of the war on some of the primary pathways through which it had long lasting In addition to SHARE data, we also use external data impacts: hunger, dispossession, the absence of a father, and sources to identify aggregate channels of war affectedness. marriage. The final section highlights our main conclusions. Since World War II affected not only countries differentially but also regions within countries, we constructed data II. SHARELIFE Data on combat operations using sources from military history (Ellis, 1994). Using maps of within-country regions for A. SHARE and Retrospective Early-life Data from each month during the war, we documented whether armies SHARELIFE engaged in battle in that place at that time. We combined these data with information about the region in which SHARE is a multidisciplinary cross-national panel inter- respondents lived during each year of the war and use this view survey on health, socioeconomic status, and social and material as one measure of individual war exposure. family networks of individuals aged 50 or over in continen- Since we analyze data over a time period of fifty years, tal Europe. The original 2004/2005 SHARE baseline in- we also have to account for country-specific economic per- cluded nationally representative samples in eleven European formance that may have affected childhood circumstances countries (Denmark, Sweden, Austria, France, Germany, differently. We therefore use GDP data, which are available Switzerland, Belgium, Netherlands, Spain, Italy, and for each European country (Maddison, 2011). We also used Greece) drawn from population registries or from multistage external data on country-specific civilian and military caus- sampling (http://www.share-project.org/). For these coun- alities associated with the war, population movements, and tries, a second wave of data collection took place in 2006, the sex ratio. Table 1 contains definitions of variables and the third wave of data collection on this panel (SHARE- derived from SHARE and SHARELIFE that we used in our LIFE) was completed in 2008. analysis in this paper. Appendix table A1 provides a paral-

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/REST_a_00353 by guest on 29 September 2021 THE EFFECTS OF WORLD WAR II ON ECONOMIC AND HEALTH OUTCOMES ACROSS EUROPE 105

TABLE 1.—VARIABLE DEFINITIONS Variable name Definition Background information Year of birth Year of birth Male Dummy ¼ 1 if respondent is male Childhood SES Unifies four measures for SES at age 10: logged number of books in household; logged number of rooms and persons in household; features in household; occupation of main breadwinner Outcome measures Childhood immunizations Dummy ¼ 1 if respondent got any vaccinations during childhood Depression Dummy ¼ 1 if respondent suffers from more than three depression symptoms in EURO-D scale Diabetes Dummy ¼ 1 if respondents has diabetes or high blood sugar Ever married Dummy ¼ 1 if respondents was ever married Heart disease Dummy ¼ 1 if respondent has heart problems (including heart attack) Height Height in cm Life satisfaction Life satisfaction on a scale from 0 to 10, with 0 very unsatisfied and 10 very satisfied Log(net worth) Logged household net worth as the sum of values from bank accounts, bonds, stocks, mutual funds, retirement accounts, contractual savings and life insurances minus liabilities Self-rated health Categorical variable for self-rated health with excellent health ¼ 5 Years of education Years of education Channels of war exposure Dad absent Dummy ¼ 1 if biological father was absent at the age of 10 Dispossession Dummy ¼ 1 if respondent reports ever being dispossessed Hunger Dummy ¼ 1 if respondent ever suffered hunger and when Persecution Dummy ¼ 1 if respondent reports ever being persecuted War variables War Dummy ¼ 1 if respondent was living in a war country during the war period War combat 0–2 months Respondent was living in a war country during the war period in a region within the country that experienced 0 to 2 months of combat War combat 3–10 months Respondent was living in a war country during the war period in a region within the country that experienced 3 to 10 months of combat

lel list of variables constructed from external data sources TABLE 2.—GDP PER HEAD RELATIVE TO US GDP PER HEAD with a documentation of the source that was used. Country 1938 1950 1973 1987 United Kingdom .98 .72 .72 .73 III. The Channels of Long-Term Effects of Germany .84 .45 .79 .82 France .72 .55 .78 .78 World War II Italy .53 .36 .63 .70 Japan .38 .20 .66 .77 This section presents descriptive data and reviews the USSR .35 .30 .36 .33 current literature on possible major channels through which Source: Harrison (1998, tables 1–10). World War II might have affected people’s lives well into their older years. The channels include future per capita the losing side—Germany, Japan, and Italy—presumably income growth of countries affected, mortality, changing reﬂecting their much larger losses in both physical and sex ratios, absence of a father, periods of hunger, migration, human capital during the war. However, by 1973 and cer- dispossession, and persecution. This section is used to moti- tainly by 1987, the European ‘‘losers’’ had higher per capita vate the rationale for analyses pursued in section IV. growth than the European ‘‘winner’’. What appears to be essential in the long term was not whether a country was on the winning or losing side, but whether it transited to A. Per Capita Income Growth democracy and open-market economies.3 The poor perfor- 4 If wars alter long-term economic growth, they would per- mance of former USSR countries illustrates that point. manently depress economic prospects of future generations. Warfare reduces capital stock through the destruction of 3 Waldinger (2012) demonstrated one microchannel on human capital, the loss of Jewish university professors in Germany due to the war. He infrastructure, productive capacity, and housing through shows that the productivity of those departments in Germany thar lost a rela- bombing and ﬁghting and results in a relocation of food and tively high share of their professors was permanently lowered, while shocks other production into military production. It obviously to physical capital due to Allied bombing had returned to their old growth path by the 1960s. More generally, to make up for investments in human destroys human capital. The question for our analysis is, capital takes years, while plants and factories can be repaired and replaced Will there be catch-up growth, or will the destruction show much more quickly. up many decades later? 4 A related issue is the impact of World War II on population growth in countries and affected cities. In spite of deaths of large numbers of civi- Based on Harrison (1998), table 2 displays GPD per lians in the war, the evidence indicates that affected cities went back to capita of some of the major countries involved in the war their former population growth paths in Western Germany and Japan relative to that of the United States at key illustrative dates. (Brakman, Garretson, & Schramm, 2004), while city growth, but also economic growth, was permanently depressed in East Germany and the The immediate impact of the war was apparently quite Soviet Union (Acemoglu, Hassan, & Robinson, 2011; Brakman et al., destructive for the countries involved, especially those on 2004).

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/REST_a_00353 by guest on 29 September 2021 106 THE REVIEW OF ECONOMICS AND STATISTICS

FIGURE 1.—WORLD WAR II CASUALTIES

B. Mortality Among European countries covered by our data, Germany and Poland bore the brunt of these casualties. In contrast and In 1939, there were about 2 billion people in the world. for comparative purposes only, American casualities in the The best estimates indicate that between 62 and 78 million European and Asian theaters combined were a bit over of them would die due to World War II—more than 3% of 400,000, the overwhelming majority of them soldiers. Simi- the world’s population. While earlier wars also resulted in larly, total deaths in the United Kingdom are estimated to be deaths of civilians, civilians were particularly heavily about 450,000, 15% of them civilians. affected by World War II, with about half of the European 5 Figure 1B displays total number of deaths by type in the casualties being civilians. Among civilian deaths, between same countries. Deaths were highly concentrated in Ger- 9.8 and 10.4 million were murdered for political or racial many and Poland, where deaths measured around 5 million reasons by the Nazi regime (Auerbach, 1992). Deaths due to in both countries. In Germany, there were almost as many the war were unequally distributed across countries, whether civilian deaths as military ones, while in Poland, civilian they were military deaths due to combat, civilian deaths, or deaths including the Holocaust are by far the dominant the Holocaust. Figure 1A displays the fraction of the 1939 ones. In many of the remaining countries in our data, deaths population who died in a large array of affected countries. due to the war are measured instead in the hundreds of thousands but still often amount to a large fraction of the prewar populations in several other countries, particularly 5 For example, in World War I, there were 16 million total deaths of which almost 10 million were military deaths. Most of the civilian deaths Austria and the Netherlands. The other European countries in that war were due to famine and disease. that stand out are those that would comprise most of the

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/REST_a_00353 by guest on 29 September 2021 THE EFFECTS OF WORLD WAR II ON ECONOMIC AND HEALTH OUTCOMES ACROSS EUROPE 107

TABLE 3.—PERCENT OF SHARELIFE RESPONDENTS WITH FATHER ABSENT AT AGE 10, BY TIME PERIOD Before After Before After Country 1939 1939–1944 1945–1949 1950–1955 1955 1939 1939–1945 1945–1949 1950–1955 1955 % Father Absenta % Hungry Austria 0.195 0.256 0.274 0.221 0.152 0.054 0.116 0.078 0.026 0.015 Germany 0.113 0.191 0.223 0.297 0.099 0.006 0.211 0.162 0.029 0.016 Sweden 0.113 0.126 0.092 0.140 0.123 0.006 0.008 0.005 0.005 0.018 Netherlands 0.074 0.083 0.102 0.073 0.044 0.005 0.116 0.008 0.001 0.003 Italy 0.044 0.103 0.068 0.069 0.046 0.032 0.123 0.053 0.027 0.020 France 0.106 0.149 0.158 0.128 0.056 0.008 0.117 0.026 0.014 0.026 Denmark 0.073 0.092 0.118 0.103 0.076 0.000 0.004 0.001 0.001 0.007 Greece 0.019 0.024 0.042 0.049 0.013 0.032 0.123 0.046 0.027 0.014 Switzerland 0.067 0.054 0.085 0.030 0.046 0.014 0.013 0.008 0.007 0.013 Belgium 0.046 0.083 0.065 0.086 0.055 0.004 0.070 0.012 0.006 0.014 Czechia Republic 0.080 0.063 0.134 0.095 0.081 0.012 0.038 0.021 0.006 0.004 Poland 0.078 0.115 0.173 0.154 0.082 0.015 0.123 0.054 0.018 0.033 Sex Ratios % Dispossession % Ever Dispossessed Before 1939 After 1945 1945–1948 Austria 0.004 0.015 0.003 0.000 0.000 0.029 0.95 0.86 - Germany 0.004 0.039 0.011 0.007 0.006 0.064 0.96 0.81 0.72 Sweden 0.000 0.002 0.000 0.000 0.001 0.003 0.99 1.02 1.02 Netherlands 0.002 0.019 0.000 0.000 0.000 0.015 0.99 0.97 - Italy 0.001 0.004 0.000 0.000 0.000 0.004 0.94 0.95 0.94 France 0.003 0.024 0.001 0.001 0.006 0.032 0.96 1.02 - Denmark 0.000 0.003 0.000 0.000 0.001 0.004 0.95 0.99 - Greece 0.014 0.011 0.001 0.000 0.001 0.014 0.92 0.99 - Switzerland 0.006 0.002 0.000 0.001 0.003 0.011 0.96 0.99 0.98 Belgium 0.007 0.035 0.003 0.000 0.002 0.036 0.99 1.02 - Czechia Republic 0.018 0.012 0.020 0.078 0.023 0.136 0.95 0.97 0.97 Poland 0.002 0.058 0.008 0.001 0.003 0.065 0.89 - - The countries are deﬁned as of 2009 when SHARELIFE data were collected. One-year refers to year at age 10. aRefers to the year when respondents were age 10. Source: SHARELIFE; calculations by authors. For sex ratios, see A1.

Soviet Union, where one in seven perished in the war, with old. Once again, the largest effects took place in the war- about 10 million military deaths and 13 million civilian ravaged countries of Austria, Germany, and Poland. In Aus- deaths. Unfortunately, data on these countries are not part tria and Germany, about one in four children lived without of the SHARE network of European countries.6 their biological fathers when they were age 10 during the war. The legacy persists into years after the war, since many who were age 10 from 1950 to 1955 had fathers who C. Sex Ratios and Absence of Father died during the war. In Germany, almost a third of those age 10 in these years were not living with their biological Mostly men died during the war, producing low male-to- father. Absent father rates fall sharply in the post-1955 female ratios in Europe after the war, as well as the absence years since these children were born after the war. We of many fathers during the respondents’ childhood years. observe war spikes in other countries as well (Italy, France, Since the male bias in deaths was concentrated among sol- Denmark, and Belgium), but the contrasts with the pre- and diers, as civilian and Holocaust fatalities were largely gen- postwar years are not as dramatic. der neutral, it is the countries in ﬁgure 1 that experienced Sex ratios before, during, and after the war are contained many military deaths that were most affected. With 3 mil- in the bottom-right half of table 3. In Germany, the sex ratio lion military deaths, the most affected country in our data dropped from 0.96 in 1939 to 0.72 men per women in the was Germany. 15 to 45 age group after the war in 1946. Thus, many The top left-hand side of table 3 shows one immediate women did not marry, and many children grew up without demographic consequence of the war by listing by country a father. Even after the war, about 4 million of the 11 mil- and period when one was age 10: the fraction of individuals lion German prisoners of war remained in captivity, and the who had a biological father absent when they were 10 years last 35,000 German soldiers returned from the Soviet Union in 1955, which compounded the problem of absent fathers 6 While we concentrate for data reasons primarily on the effects of (Wehler, 2008). World War II in continental Western Europe, the war’s impact was just as stark in the Asian theater. The two countries most directly affected in terms of number of casualities were Japan and China. About 2 million D. Hunger Japanese soldiers died in the war alongside up to 1 million Japanese civilians—about 4% of the prewar Japanese population. The total number of deaths in China is believed to range between 10 and 20 million, with more One channel by which World War II might have affected than 70% being civilians. long-run adult health and SES outcomes is hunger. The war

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/REST_a_00353 by guest on 29 September 2021 108 THE REVIEW OF ECONOMICS AND STATISTICS

FIGURE 2.—PERCENTAGE OF SHARE RESPONDENTS SUFFERING FROM HUNGER:WAR VERSUS NONWAR COUNTRIES

caused several severe hunger crises, which led to many the end of the German occupation in May 1945. The Dutch casualties and may have had long-term effects on the health famine has been extensively studied because it affected an of survivors. For example, since the beginning of the Ger- otherwise well-nourished population at a very specific time man occupation in Poland, the nutritional situation of the and region. Individuals exposed to this famine in utero are non-German population was poor. The average caloric shown to suffer from cognitive and mental problems and intake for the Polish population was about 930 calories in addiction (Neugebauer et al., 1999; Rooij et al., 2010), dia- 1941. The situation was worst in the Warsaw Ghetto, where betes, and coronary heart disease, and they also perform average food rations were limited to about 186 calories per worse regarding anthropometric and socioeconomic indica- day in 1941. tors (Almond & Currie, 2011). Similarly, in the fall and winter of 1941–1942, Greece Germany suffered from hunger between 1945 and 1948 was struck by a severe famine. with about 100,000 to when the food supply from occupied countries ceased. In 200,000 deaths (Hionidou, 2006). During the war, Greece the U.S. occupation zone, the Office of Military Govern- was under Bulgarian, German, and Italian occupation. The ment for Germany established a goal of 1550 calories per famine was mainly caused by three factors: occupiers day in 1945, but in the first months of occupation, this goal imposed a naval blockade, prices to farmers were fixed at often could not be met. There were regions where average such low levels that they were not willing to market their calories per day were around 700 (Gimbel, 1968). Death products, and mobility between different regions of the rates raised by a factor of 4 for adults and 10 for infants dur- country was reduced due to occupation. The nutritional ing this period. With a good harvest and currency reform in situation returned to acceptable levels toward the end of June 1948, nutritional shortages were overcome (Zink, 1942. Neelson and Stratman (2011) use cohort data to show 1957). that undernourishment of children who were 1 or 2 years Figure 2 demonstrates that hunger episodes during the old at the time of the famine had a significantly lower prob- war were much more severe in war countries than in coun- ability of being literate or to complete upper secondary edu- tries that did not participate in the war. We also see the cation. great amount of diversity in periods of hunger within war A combination of a food blockade and a harsh winter led countries. Hunger is more common in regions where com- to a severe hunger crisis in the winter of 1944–1945 in the bat took place within war countries. Finally and not surpris- Netherlands. About 20,000 deaths, mainly among elderly ingly, the experience of hunger was far more common men, are attributed to this famine. The famine ended with among those of low socioeconomic background as a child.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/REST_a_00353 by guest on 29 September 2021 THE EFFECTS OF WORLD WAR II ON ECONOMIC AND HEALTH OUTCOMES ACROSS EUROPE 109

With respect to hunger, our analysis shows that the indivi- FIGURE 3.—TOTAL INFLOW AND OUTFLOW OF POPULATION, 1939–1947 dual-level reports in SHARELIFE match well with historical information on the timing and location of hunger episodes we collected from historic sources. To illustrate, in ﬁgure 2 the Greek hunger spike occurred in 1941–1942, the Dutch in 1944–1945, and the German in 1946–1947.

E. Dispossession, Persecution, and Migration SHARELIFE documents the extent of the experience of dispossession of property linked to World War II and its aftermaths. Dispossession was often associated with persecution and resulted in geographic displacement of populations during and immediately after the war. A further advantage of SHARELIFE is that we can observe where and when individuals moved during their lifetimes, including the wartime period. There were three main periods when people were forced to flee their homelands. During the war, millions of Jews, but also opponents of the Nazi regime, were often sent to concentration camps and murdered there. Second, the end of the war was associated with dramatic border changes in Eastern Europe, which induced millions of individuals to leave their places of residence and flee to other parts of Eur- ope. The Soviet Union annexed territory from some of its neighboring countries, inter alia from Czechoslovakia, Ger- many, and Poland. Poland in turn received one part of prewar Germany in compensation. Poles, having lost their homes in the part occupied by the Soviet Union, were moved to the new part, so Poland and, with it, millions of living in Germany and Poland, dispossession happened people were moved westward. more frequently during the war period, while they happened Figure 3 shows inflows and outflows of populations dur- after the war in Czechoslovakia. Dispossessed individuals ing and after the end of World War II into the new states in in our sample are proportionally more likely to have been their new borders. Germany lost about one-quarter of its born outside the current borders of their country. Analyzing territory. About 2 million people have been estimated to countries of origin, many of them came from Eastern Eur- have died on the flight. After the war, the remaining terri- ope; thus they probably lost their property with the large tory of Germany was divided into four occupational zones. wave of nationalizations after the war: Not surprisingly, it About 4 million people fled from the approaching Soviet is the foreign born living in our SHARE countries who were armies to the British and U.S. zones where the occupation most likely to be dispossessed. was less severe. In Germany, destroyed cities had to accom- modate millions of ethnic Germans from other parts of Eur- IV. World War II and Individual Outcomes: Analysis of ope. A further wave of dispossessions happened in Eastern SHARELIFE Data countries after the war, when private property was nationalized in the socialist and communist economies. Even in Based on the descriptive data and review in the prior sec- France, there was a wave of nationalizations at the end of tion, we find enormous variation even among war countries the war. Mainly banks, energy, and transport firms were in the immediate impact of World War II. Long-term eco- nationalized, but there were also some expropriations as a nomic or population growth rates seem unlikely to be a pri- penalty for cooperation with the Nazi regime. mary pathway through which the war’s influence took The bottom left-hand side of table 3 displays disposses- place. Instead, changing gender ratios induced by differen- sion rates in our SHARE countries by time period, with the tial male mortality in the war appear to be a more plausible final column indicating the percent ever dispossessed. Fig- pathway operating through both absence of fathers and dif- ure 4 complements the data in table 3 by showing the per- ficulties faced by women in marrying. Hunger and immedi- centage of dispossessed individuals in SHARELIFE for the ate and long-term stress created by battles, dispossession, foreign and native-born populations. In the Czech Republic, and persecution would also appear to be plausible pathways Germany, and Poland, more than 5% of respondents experi- that could affect adult health, both physical and mental, and enced dispossession during their lifetime. For respondents our later-life measures of adult SES.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/REST_a_00353 by guest on 29 September 2021 110 THE REVIEW OF ECONOMICS AND STATISTICS

FIGURE 4.—DISPOSSESSION OF POPULATION IN WORLD WAR II

A. Measures of War Exposure TABLE 4.—NUMBER OF OBSERVATIONS AVAILABLE IN SHARELIFE BY COUNTRY Country War ¼ 0 War ¼ 1 Total To analyze long-term impacts of World War II on health and economic outcomes, we use the fact that different coun- Austria 146 565 711 Germany 450 1,001 1,451 tries in Europe and different people in those counties were Italy 863 1,470 2,333 differentially affected by the war at different points in time. Czechia 723 925 1,648 To study the effects on adult outcomes, we use two indica- Greece 1,149 1,482 2,631 Poland 819 758 1,577 tors of being affected by World War II: (a) that one lived in Belgium 1,026 1,480 2,506 a war country during the war period and (b) that one was France 793 1,105 1,898 exposed to combat in the area within a country in which Netherlands 883 1,069 1,952 Denmark 1,927 0 1,927 one lived during the war. Our first measure essentially cre- Sweden 1,639 0 1,639 ates a war dummy equal to 0 for everybody in a nonwar Switzerland 993 0 993 country (Denmark, Switzerland, and Sweden) and for Total 11,411 9,855 21,266 The countries are defined as of 2009 when the SHARELIFE data were collected. Native born only. everybody born after the war period no matter what country War ¼ 1 means that the respondent was living in a war country sometime during World War II. they lived in. Alternatively, it is equal to 1 for everybody Source: SHARELIFE; calculations by authors. alive in a war country (Austria, Belgium, Czech Republic, France, Germany, Greece, Netherlands, and Poland) during the war period. The war period ends in 1945 for all war this variable was to test whether actual exposure to combat countries and includes 1946 to 1948 in Germany and Aus- was an important mechanism for the war effects that we tria, when they were under Allied occupation. For these estimate below. countries, the war period ended with the currency reform in Table 4 provides the list of SHARE countries that are Germany in 1948. Individuals could certainly have been part of our analysis with the sample sizes of those SHARE- affected by the war even if they were born after the war, but LIFE respondents who experienced the war and those who the channels we emphasize in this paper—combat, hunger, had no direct experience of war. We did not include Spain dispossession, persecution, and the absence of a father— in our analysis since Spain experienced a civil war in the were more likely to have affected those who lived during late 1930s, so a distinction between whether Spain was a the war. war country or not is ambiguous. The results were not sig- Our second war measure involves constructing a variable nificantly different if Spain was included. indicating whether there were combats and how many combats occurred in the region within the country in which the B. Microlevel Regressions of Adult Health and SES individual lived during the war. Thus, in the war countries, Outcomes we created two dummy variables based on the number of months of exposure respondents had to combat in the place We next turn to our statistical modeling of whether indi- they lived during the war: 0 to 2 months of exposure to viduals’ experiences during World War II predict their combat and 3 or more months of exposure. The purpose of health and socioeconomic status in their later adult life. For

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/REST_a_00353 by guest on 29 September 2021 THE EFFECTS OF WORLD WAR II ON ECONOMIC AND HEALTH OUTCOMES ACROSS EUROPE 111

all of our later-life health and SES outcomes and channel FIGURE 5.—LATER-LIFE OUTCOME DIFFERENCES BETWEEN WAR AND NONWAR outcomes, our estimating equation takes the form COUNTRIES BY PERIOD OF BIRTH

Yitc ¼ b1 waritc þ b2 malei þ kt þ gc þ eitc ð1Þ

where Yitc is the later-life outcome of respondent i born in year t and living in country c. Male indicates a respondent was male. War is one of our two measures of war exposure outlined above, which vary by country (or region within a country) and year of birth. Because there may be unmea- sured country and year effects associated with these outcomes, lt is a full set of year-of-birth dummies and Zc is a full set of country dummies. eitc is a random error term. Since error terms within country and within year may be correlated, we used the cluster option in STATA. Our principal interest is to obtain estimates of b1—the war effect in addition to birth-year and country effects. We estimate reduced-form models using our two war variables on later adult life health and SES outcomes and the principal channels of war. We consider several adult dependent variables all measured in 2009, the year of SHARELIFE. Health outcomes include prevalence of diagnosed diabetes and heart disease, body height in centimeters (a summary measure of early-life health conditions), whether an individual is depressed using a dummy variable for the presence of at least four symptoms on the EURO-D scale, and self- reported health status. Self-reported health status is re- corded on a scale of excellent, very good, good, fair, and poor, which we have translated to a scale from 1 to 5, with 5 the best health status. Our adult SES and economic outcomes include log of household net worth, whether the individual was ever married, and life satisfaction in 2009. SHARE respondents are asked, ‘‘On a scale from 0 to 10 where 0 means completely dissatisfied and 10 means completely satisfied, how satisfied are you with your life?’’ We model this outcome as a score from 0 to 10. We have two education measures in SHARE. The first is obtained from baseline SHARE in 2004 and, in an attempt to make the education variable comparable among individuals in the same country, assigns a standardized year for each education value. For example, university graduates in a country would be assigned a 16. The second education variable is available in the second SHARE wave and is equivalent to the number of years spent in education. We use the second measure because Poland and the Czech Republic were not part of baseline SHARE, and for those two countries the first measure is not available. However, we hypothesize that World War II may have disrupted edu- with time period of birth using three subsets of countries— cation for many respondents and resulted in a longer time Germany and Austria combined, other war countries, and to complete a given level of education. To test that hypoth- the nonwar countries. These outcomes are each expressed esis for the subsample of respondents who have both mea- as the difference between each of the first two kinds of war sures of education from the second and first SHARE waves, countries minus the outcome in the nonwar countries. For we estimated a model that amounts to the difference all three outcomes, the outcomes deteriorate relative to the between the two education measures (the second-wave edu- nonwar countries for those born at a time they would cation minus the first-wave education variable). experience war. Figure 5 displays the association of three of our key out- Table 5 summarizes results obtained for adult health outcomes—education, self-reported health, and depression— comes and table 6 for adult SES outcomes. We present

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/REST_a_00353 by guest on 29 September 2021 112 THE REVIEW OF ECONOMICS AND STATISTICS

TABLE 5.—ADULT HEALTH OUTCOMES ASSOCIATED WITH WORLD WAR II Diabetes Heart Height Depression Self-Reported Health Variables (1) (2) (3) (4) (5) A. War variable War 0.026*** 0.014 0.196 0.058*** 0.094*** [0.009] [0.010] [0.179] [0.014] [0.034] Male 0.010** 0.045*** 11.579*** 0.170*** 0.115*** [0.004] [0.004] [0.106] [0.006] [0.015] Observations 21,228 21,228 21,115 21,266 21,254 R2 0.022 0.061 0.510 0.073 0.148 B. Combat variable War combat 0–2 months 0.030*** 0.011 0.138 0.047*** 0.091*** [0.009] [0.010] [0.205] [0.015] [0.036] War combat 3 or more months 0.018* 0.018* 0.308 0.078*** 0.098*** [0.010] [0.012] [0.221] [0.016] [0.037] Male 0.010** 0.045*** 11.576*** 0.169*** 0.114*** [0.004] [0.005] [0.086] [0.006] [0.015] Observations 21,212 21,212 21,099 21,250 21,238 R2 0.022 0.061 0.510 0.073 0.148 OLS regressions include both country dummies and birth-year dummies. Robust standard errors in brackets allow for correlation at year/birth level. ***p < 0.01, **p < 0.05, and *p < 0.1.

TABLE 6.—ADULT ECONOMIC OUTCOMES ASSOCIATED WITH WORLD WAR II Education Ln Net Ever Life Education Difference Worth Married Satisfaction (1) (2) (3) (4) (5) A. War variable War 0.274** 0.362*** 0.076 0.032*** 0.306*** [0.127] [0.126] [0.057] [0.008] [0.051] Male 0.964*** 0.084 0.224*** 0.017*** 0.212*** [0.060] [0.056] [0.022] [0.005] [0.025] Currently married 0.859*** [0.045] War male 0.024*** [0.007] Observations 19,572 11,355 20,341 21,266 19,549 R2 0.276 0.145 0.356 0.011 0.128 B. Combat variable War combat 0–2 months 0.215 0.406*** 0.097* 0.030*** 0.242*** [0.134] [0.129] [0.058] [0.008] [0.054] War combat 3 or more months 0.398*** 0.264** 0.029 0.036*** 0.436*** [0.134] [0.154] [0.065] [0.009] [0.058] War combat 0–2 months Male 0.029*** [0.008] War combat 3 or more months Male 0.020*** [0.007] Male 0.965*** 0.085 0.223*** 0.017*** 0.211*** [0.060] [0.056] [0.022] [0.005] [0.025] Currently married 0.860*** [0.045] Observations 19,556 11,347 20,327 21,250 19,533 R2 0.268 0.145 0.356 0.011 0.129 OLS regressions include both country dummies and birth year dummies. Robust standard errors in brackets allow for correlation at year/birth level. *** p < 0.01, ** p < 0.05, and * p < 0.1.

regressions in the A panels that use only the aggregate war tion), missing values are on the order of 1 in 1,000 observa- exposure measures and in panel B the measure that distin- tions. guishes between very limited exposure to combat (two Consistent with the literature, men have higher levels of months or less, including 0) or more extensive combat adult diabetes and heart disease, lower levels of depression, exposure (three or more months) with the left-out category and report themselves in better subjective health than being not exposed to war at all. In terms of right-hand-side woman do (Banks, Muriel, & Smith, 2010; Smith, 2007). variables, there are no missing values for gender. If the Our principal concern involves estimates for aggregate war outcome in any particular model is missing, this ob- and combat variables. Living in a war country during the servation was not included in that speciﬁc model. Missing period of World War II is consistently statistically signiﬁ- values in our outcomes are relatively rare. In terms of main cantly associated with higher levels of adult diabetes, being channels (dad absence, dispossession, hunger, and persecu- more depressed, and reporting one’s subjective health as

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/REST_a_00353 by guest on 29 September 2021 THE EFFECTS OF WORLD WAR II ON ECONOMIC AND HEALTH OUTCOMES ACROSS EUROPE 113

worse. Being in a war country during the war increased the TABLE 7.—MEAN NUMBER OF CHILDREN PER WOMEN probability of diabetes in later life by 2.6 percentage points Median SES split Prewar During War Postwar and depression by 5.8 percentage points while decreasing Low SES 1.28 1.43 2.47 self-reported health by 9.4 percentage points. These High SES 1.11 1.25 2.25 increases are all high relative to baseline rates (see table A2). Estimated effects on heart disease and height are not statistically significant. on postwar savings behavior and trends in asset prices. The Panel B of table 5 displays results for months of combat war combat models in the panel B of table 6 produce exposure variables: number of months of exposure respon- roughly similar results in direction and magnitude of these dents had to combat in the place they lived during the war outcomes. in war countries using zero to two months of exposure to One purpose of our combat variables was to test whether combat and three or more months of exposure to combat. the actual exposure to combat was an important mechanism These results basically parallel those obtained for the war for the war effects that we estimate above. With the sole variable in both direction and magnitude: those with combat exceptions of adult depression (table 5) and life satisfaction exposure were more likely to have diabetes as an adult, (table 6), the estimated magnitude of the worse adult SES were in worse self-reported health, and were more likely to and health outcomes appears to be about the same among be depressed. The results are weaker for heart problems, those with large or small exposures to actual combat.8 This although three or more months of combat exposure suggests that experiencing combat and battles close by to increases the likelihood of heart disease as an adult and is where you lived during the war are not the primary mechan- statistically significant at the 10% level. ism by which these war effects operate. The exceptions are Table 6 repeats the same type of models for adult eco- of interest since it seems reasonable that frequent exposure nomic outcomes in 2009. Not surprisingly for these genera- to combat is associated with adult depression and lower tions, compared to women, men achieve more years of levels of life satisfaction as the vivid memories of that schooling, have higher net worth, are less likely to marry, experience persist into adulthood. and have higher levels of life satisfaction, common findings in the literature. Our measure of war exposure is strongly associated with all of these SES outcomes, except ln net C. Selection Effects worth. Those in a war country during the war achieved As in any such analysis, there are issues of possible about three-tenths of a year less education7 and achieved selection effects due to fertility, mortality, and migration lower levels of life satisfaction (about a third of a point that may have biased our estimates. The concern with selec- lower relative to a mean of 7.6) as older adults. The educa- tive fertility is that high-SES mothers reduced their fertility tion difference model suggests that war makes respondents more during the war, which on average would lead to less take longer (a third of a year) to reach a given level of edu- healthy babies. SHARE does not contain variables on education. Similarly, this exposure to war reduced the probabil- cation of parents, so we used instead our measure of child- ity of women being ever married (about 3 percentage hood SES, acknowledging its possible endogeneity. We points) but not the marriage probability for men, consistent examined fertility in SHARE in three time periods by SES: with the relative scarcity of men due to war. In contrast, ln prewar (before 1939), during the war (1939–1945), and household net worth is not associated with the wartime postwar (after 1945). Childhood SES was split at the med- experience, suggesting that this outcome mainly depends ian (see table 7). In all three periods, fertility is highest in the low-SES 7 Ichino and Winter-Ebmer (2004) compare educational outcomes from cohorts affected by the war in Austria and Germany to cohorts in Switzer- groups. But differential changes by SES in fertility across land and Sweden, using the main economic data sets with information on these three time periods do not seem large enough to be pro- education and earnings in the countries. They find that the loss of school- ducing our results. Comparing prewar and during–war peri- ing is about a fifth of a year compared to the following cohort. They suggest that the mechanisms are closing down of religious schools, absence ods, there was about a 0.14 increase in fertility for both low- of teachers due to the war, absence of students due to escaping bombing, SES and high-SES groups. Similarly, comparing postwar to and destruction of schools. Akbulut-Yuksel (2009) uses GSOEP to iden- during-war periods, average fertility rose by about one child tify the effects of destruction of German cities through bombing on schooling. She finds that destruction caused children to attain 0.4 fewer in both SES groups. Moreover, when we added childhood years of schooling. Her estimates suggest that this schooling reduction is SES measures to our models, which should partly control mainly due to physical destruction of schools and the absence of teachers. for any selective fertility associated with the war, our esti- Jurges€ (2011) uses the Micro-Census to analyze impacts of nutritional shortages in Germany on educational outcomes. He estimates a drop in mates of the long-term effects of war did not change much. educational achievements (having more than basic education) of about 5 Individuals in our analytical sample are those still alive percentage points for a baseline risk of about 30%. His suggested pathway in 2009, so they are a selected sample of the population that is nutritional deprivation in utero. Our result of about a third of a year of schooling lies between the estimates of Ichino and Winter-Ebmer and Akbulut Yuksel, but is lower than those of Jurges.€ However, we include a 8 For the two exceptions—depression and life satisfaction—the effect different set of countries, and not all of them were equally strongly of three or more months of combat greater than zero to two months is sta- affected by World War II as Germany was. tistically significant at the 1% level.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/REST_a_00353 by guest on 29 September 2021 114 THE REVIEW OF ECONOMICS AND STATISTICS

TABLE 8.—AGE OF DEATH OF FATHERS were also asked whether they had ever been victims of per- Low SES High SES secution because of their political beliefs, religion, national- Country Before 1946 After 1945 Before 1946 After 1945 ity, ethnicity, sexual orientation, or background. Unfortu- nately, no time period for that persecution was asked. Nonwar 73.76 72.89 73.73 72.95 War 71.22 70.87 70.67 70.36 Finally, the absence of the father is defined as the absence of the biological father at the age of 10. About 8% of our respondents experienced a period of hunger, 9% lived without their father at age 10, and 5% suffered from persecution experienced World War II. To the extent that those more and dispossession, respectively. We also included in our affected are less likely to have survived, our results should models an additional possible pathway: whether a respon- understate the full effects of war on long-term health and dent received immunizations as a child. SES outcomes. A more complicated issue concerns differ- Table 9A shows how microchannels are related to the ential mortality by SES induced by the war. If mortality due experience of war. Males are both more likely to suffer to the war was much higher in low-SES groups (whose from hunger and to be persecuted. The latter is what we health would have been worse anyway), we would further expected given that mainly men were politically active dur- understate health effects of war. We examined data on age ing this period. Having experienced World War II increases of death of father by SES by whether one lived in a war or the likelihood of experiencing hunger by about 8 percentage nonwar countries and by whether those we surveyed experi- 9 points, dispossession and persecution by 1 percentage enced the war as a child (born before 1946) (table 8). Once point, the absence of a father by 2 percentage points. These again, dividing SES at the median we found the following estimates are large relative increases given baseline risks. for the mean age of death of father. Those born after 1946 For example, the probability of experiencing hunger is who did die should be younger, but the key comparison is doubled by war exposure, and the probability of an absent differentials by SES. dad is increased by 25% in relative terms. The experience For nonwar countries, we find in comparing pre- and of war was associated with a lower probability of immuni- post-1945 that the age of death of father decreased by 0.8 of zation as a child, which is unsurprising given that this was a year in both low- and high-SES groups. Using the same wartime. This immunization result may be a pathway comparison, the age of father fell by 0.4 of a year in the war through which adult health eventually suffers. countries, but this was approximately the same for the low- In table 9B our interest lies in whether the experience of and high-SES groups. Once again, this degree of selection combat is the mechanism that leads to war effects. Once does not seem large enough to be driving our results. again, the size of these estimates is very similar to those Because of population shifts, especially inflows docu- obtained by the country wartime variable. The experience mented in figure 3, we confined our analysis to the native- of hunger and absence of the father is somewhat stronger born in each country. Among countries in our data, figure 3 for our respondents who lived in a region strongly affected shows that outflows were significant only in the Czech by combat (three to ten months of combat) than for those in Republic, Poland, and Germany. Since it was not encour- regions with none or only a mild experience of combat. aged by receiving countries, migration during and after the However, the differences are not large. In fact, we expect war was quite difficult in Europe. But there was some the death of men during wartime to not necessarily happen migration, and one must allow for the possibility that selec- in their region of residence. Persecution is related to war tive migration may influence our estimates on war effects especially for these three countries. Of course, people could have temporarily left combat areas as combat was taking 9 Regarding the influence of hunger on later-life outcomes, we do not present structural estimates of the influence on later-life outcomes as there place but stayed in the same country, which should lead to are no suitable instruments for the whole of Europe. These types of esti- an understatement of combat effects. mations are possible for a smaller set of countries. Van den Berg, Pinger, and Schoch (2011) use hunger periods caused by World War II for Greece, Germany, and the Netherlands as instruments to establish causal effects of undernutrition on hypertension and adult height. For Germany, D. Models of Channels of War we collected data on monthly caloric rations in regions where respondents live. We see large drops in calories toward the end of the war and in occu- We next turn to our estimates of how the micropathway pation zones, with the French and Soviet zones hit hardest. When we channels we have highlighted—hunger, dispossession, per- regress our health outcomes on average calories available between ages 0 and 16 in respondents’ region (again controlling for gender and year of secution, and the absence of father—are related to the birth), we find that an increase of 1000 kcal per day decreases the chance experience of World War II. SHARELIFE respondents of suffering from diabetes by 14.3 percentage points, increases self - were asked, ‘‘Looking back on your life, was there a period reported health by 0.7 points, and increases height by 3.3 centimeters. When we distinguish different age groups (0–4, 5–10, and 11–16), we see of time during which you were hungry?’’ If the answer was the strongest results for the 0–4 group and impacts on adult depression. yes, they were asked when this occurred. Individuals in This suggests that hunger analysis should be seen not only as operating SHARE are also asked ‘‘whether they or their family were through nutrition-related outcomes such as adult height but also and equally through adult outcomes such as depression. Our effects on height ever dispossessed of any property as a result of war or per- are similar to Van den Berg et al. (2011), who find an effect of between 3 secution,’’ and if yes, the date of that dispossession. They and 6 centimeters.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/REST_a_00353 by guest on 29 September 2021 THE EFFECTS OF WORLD WAR II ON ECONOMIC AND HEALTH OUTCOMES ACROSS EUROPE 115

TABLE 9.—CHANNELS OF WAR ASSOCIATED WITH WORLD WAR II (1) (2) (3) (4) (5) Hunger Dispossession Persecution Dad Absent Childhood Immunizations A. War variable War 0.077*** 0.013*** 0.014** 0.022** 0.044*** [0.011] [0.005] [0.005] [0.009] [0.007] Male 0.014*** 0.004 0.010*** 0.003 0.001 [0.003] [0.003] [0.003] [0.004] [0.003] Observations 21,240 21,239 21,240 20,906 21,005 R2 0.077 0.047 0.020 0.032 0.079 B. Combat variable War combat 0–2 months 0.071*** 0.013*** 0.017*** 0.017* 0.044*** [0.012] [0.005] [0.005] [0.009] [0.006] War combat 3 or more months 0.091*** 0.012** 0.008 0.033*** 0.043*** [0.012] [0.005] [0.006] [0.010] [0.006] Male 0.013*** 0.004 0.010*** 0.003 0.001 [0.003] [0.003] [0.003] [0.004] [0.003] Observations 21,225 21,225 21,225 20,892 20,993 R2 0.077 0.047 0.020 0.033 0.079 OLS regressions include country dummies and year dummies. Robust standard errors in brackets allow for correlation at year/birth level. ***p < 0.01, **p < 0.05, and *p < 0.1.

per se, but not necessarily to an increased experience of for experiencing war and for childhood SES being either combat. Thus, combat does come with an increased likeli- low or middle class. To identify the distributional effects of hood of hunger as, for example, was the case in the Dutch war, we include a full set of interactions of the war with hunger winter. It can be due to other aspects of war, as was childhood SES. Once again, the results obtained are very the case for the Greek and German experience with hunger similar whether we use the war country variable or our during the war. Also, combat led to local deaths of the civi- combat variable, so table 10 displays only the results for the lian population, but military casualties and the deaths of war variable. The outcomes modeled are the same as those fathers often occur far from the families affected by it. in tables 5 to 9: adult health, adult SES, and channels of war. E. The Uneven Consequences of War We first discuss main effects of childhood SES. Com- pared to those in the high childhood SES group, those in the In addition to the models already summarized, we inves- lowest one have higher levels of adult diabetes (3.2 percen- tigated whether consequences of experiencing World War tage points), are smaller in stature as adults (1.8 centi- II vary by respondents’ socioeconomic status (SES) as a meters), experience higher levels of adult depression (2.5 child by estimating models that included interactions of war percentage points), and self-rate themselves in worse adult variables with childhood SES. Childhood SES is an index subjective health. The middle childhood SES group consis- generated by factor analysis (Mazzonna, 2011). SES unifies tently lies between the bottom and top in terms of these four measures for SES during childhood at age 10: logged adult health outcomes. These results conform to the general proportion of number of rooms in the household and per- finding in the literature that childhood economic circum- sons living in that household; logged number of books in stances are very predictive of later-life adult economic and the household; features in the household, namely, warm health outcomes (Currie, 2009; Case, Lubotsky, & Paxson, water, cold water, fixed bath, toilet inside, central heating; 2002; Smith, 2009b). Similarly, in accordance with the lit- and occupation of main breadwinner. For our analysis, we erature, higher childhood SES is associated with much divide childhood SES status into three terciles and label higher levels of adult education, net worth, and life satisfac- those terciles low, middle, and high. tion, another indication of the strong economic transmission We separate our analysis of distributional consequences across generations in these outcomes. The more novel of World War II from our main analysis since we recognize results are in the third panel of table 10, which deals with that childhood SES may partly be endogenous to the war. the channels of war. The probability of being dispossessed Given the destructive scale of the war, that including bomb- was highest in the high-childhood-SES group, not surpris- ing that sometimes destroyed civilian homes and led to the ing as there was more to capture. Persecution was also high- movement of men into the military, the possibility of such est in the high-SES category, while obtaining childhood endogenity is clearly an important caveat to keep in mind. immunizations was highest in the lowest-SES category. We did reestimate all models in tables 5 to 9 with these Absent fathers were not strongly differentiated across SES dummy variables for childhood SES terciles included, and categories. our estimates of the war barely changed. Finally, we examine differences in associations with war Our distributional results are contained in table 10. All by childhood SES categories. For childhood SES by World models continue to contain country and year-of-birth dum- War II interactions among the health variables, we find the mies and a dummy for male. We include both main effects negative health effects to be either neutral by SES cate-

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/REST_a_00353 by guest on 29 September 2021 116 THE REVIEW OF ECONOMICS AND STATISTICS

TABLE 10.—WAR INTERACTION WITH CHILDHOOD SES MODELS Health Outcomes Diabetes Heart Height Depression Self-Reported Health War 0.021** 0.006 0.263 0.045** 0.057 SES low 0.032*** 0.007 1.835*** 0.025** 0.342*** SES middle 0.014*** 0.005 0.852*** 0.002 0.137*** War SES low 0.003 0.028** 0.123 0.014 0.032 War SES middle 0.010 0.024** 0.118 0.017 0.098*** SES Outcomes Education Education Difference Ln Net Worth Ever Married Life Satisfaction War 0.044 0.021 0.082 0.056*** 0.171*** SES low 3.438*** 0.017 0.684*** 0.008 0.333*** SES middle 2.107*** 0.086 0.265*** 0.003 0.094*** War SES low 0.156 0.572*** 0.078 0.038*** 0.229*** War SES middle 0.390*** 0.316** 0.094* 0.027** 0.168*** Male 0.013 War Male 0.055*** War SES low Male 0.046** War SES mid Male 0.032* SES low Male 0.012 SES mid Male 0.003 Channels Hunger Dispossession Persecution Dad Absent Immunizations War 0.072*** 0.023*** 0.022*** 0.020* 0.024*** SES low 0.002 0.012*** 0.011*** 0.007 0.011*** SES middle 0.002 0.011*** 0.010** 0.002 0.006* War SES low 0.020* 0.009 0.013* 0.016 0.039*** War SES middle 0.008 0.015** 0.008 0.017* 0.018** ***p < 0.01, **p < .05, and *p < 0.0.

gories or that the negative health effects are concentrated those experiencing the war. To conduct this analysis, we use on the middle class as in the summary measure of self- new data, SHARELIFE, that record not only adult outcomes reported health or concentrated in the middle and lower in 2009 but also retrospective data for salient aspects of the classes as with heart disease, possibly reflecting the role of wartime experiences of respondents. We augment these data lifetime stress with that disease. with historical information on how the war affected indivi- In contrast, we find very strong interactions of a negative duals differently over time and across regions. Our data middle-class war interaction for many of our adult SES out- allow us to analyze which types of individuals were most comes: education and ln net worth. Life satisfaction decre- affected and by which channels. ments associated with the war were concentrated on the Our analysis shows that experiencing war increased the lower and middle classes. In terms of being ever married, probability of suffering from diabetes, depression, and, with the negative effects of the war were highest on the highest- less certainty, heart disease, so that those experiencing war SES women and the lowest-SES men. A summary of health or combat have significantly lower self-rated health as and SES outcomes does suggest that the middle class suf- adults. Experiencing war is also associated with less educa- fered more due to the war, with the lower class next in line. tion and life satisfaction and decreases the probability of Finally, the length of time it takes to achieve a given level ever being married for women while increasing it for men. of education due to war expands the most for the low and We also analyze pathways through which these wartime middle classes compared to the upper class. effects took place and found strong effects for hunger, dis- The bottom panel of table 10 shows that some pathways possession, persecution, childhood immunizations, and hav- through which war operates are concentrated among the ing an absent father. While a war of the magnitude of poorest households (hunger and immunizations present for World War II affected all social classes to some degree, our the middle class), some are concentrated among middle evidence does suggest that the more severe effects were on class (dad absent), and some among the highest-SES house- the middle class, with the lower class right below them in holds (dispossession). Persecution was focused on the mid- size of impact. dle and upper classes. This paper highlights advantages of having life histories in prospective studies such as SHARE. Population-based V. Conclusion economic panels are relatively recent, but combining them with life histories covering salient past personal and macro- In this paper, we present a microanalysis of the effects of events opens up many new research opportunities, of which World War II on some key SES and health outcomes of World War II is only one illustration. This is especially so in

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/REST_a_00353 by guest on 29 September 2021 THE EFFECTS OF WORLD WAR II ON ECONOMIC AND HEALTH OUTCOMES ACROSS EUROPE 117

Western Europe, where the political and economic history Kulischer, Eugene M., Europe on the Move: War and Population Changes, 1917–1947 (New York: Columbia University Press, of the past four decades is particularly rich and varied. 1948). Maddison, Angus, ‘‘Historical Statistics: Statistics on World Population, REFERENCES GDP and Per Capita GDP,’’ 1-2008. (2011). http://www.ggdc.net /MADDISON/oriindex.htm. Akbulut-Yuksel, Mevlude, ‘‘Children of the War: The Long-run Effects Mazzonna, Fabrizio, ‘‘The Long-Lasting Effects of Family Background: of Large-Scale Physical Destruction and Warfare on Children,’’ A European Cross-Country Comparison,’’ MEA discussion paper IZA discussion paper 4407 (2009). 245-11 (2011). Acemoglu, Daron, Tarek Hassan, and James Robinson, ‘‘Social Structure Neelsen, Sven, and Thomas Stratman, ‘‘Effects of Prenatal and Early Life and Development: A Legacy of the Holocaust in Russia,’’ Quar- Malnutrition: Evidence from the Greek Famine,’’ Journal of terly Journal of Economics 126:2 (2011), 895–946. Health Economics 30:3 (2011), 479–488. Almond, Douglas, and Janet Currie, ‘‘Killing Me Softly: The Fetal Ori- Neugebauer, Richard, Hans Wijbrand Hoek, and Ezra Susser, ‘‘Prenatal gins Hypothesis,’’ Journal of Economic Perspectives 25:3 (2011), Exposure to Wartime Famine and Development of Antisocial Per- 153–172. sonality Disorder in Early Adulthood,’’ JAMA 282:5 (1999), 455– Auerbach, Hellmuth, ‘‘Opfer der nationalsozialistischen Gewaltherrschaft,’’ 462. in Wolfgang Benz (ed.): Legenden, Lugen,€ Vorurteile. Ein Wo¨rter- Overman, Ru€u,€ Deutsche Militaerische Verluste im 2 Weltkrieg (Munich: buch zur Zeitgeschichte (Munich, Dtv, 1992). Oldenburg Verlag, 1999). Banks, James. Alastair Muriel, and James P. Smith, ‘‘Disease Prevalence, Putzger, Friedrich Wilhelm, Historischer Weltatlas (Biele felid: Velhagen Disease Incidence, and Mortality in the United States and in Eng- & Klasing, 1963). land,’’ Demography 47 (2010), S211–S231. Rooij, Suzanne R. de, Hans Wouters, Julie E. Yonker, Rebecca C. Painter, Brakman, Stephen, Harry Garretsen, and Marc Schramn, ‘‘The Strategic and Tessa J. Roseboom, ‘‘Prenatal Undernutrition and Cognitive Bombing of German Cities during World War II and Its Impact on Function in Late Adulthood,’’ PNAS 107:39 (2010), 16881– City Growth,’’ Journal of Economic Geography 4:2 (2004), 201– 16886. 218. SHARE Release Guide 2.5.0 for Waves 1 and 2 (2011). http://www.share- Case, Anne, David Lubotsky, and Christina Paxson, ‘‘Economic Status project.org/sharelife/. and Health in Childhood: The Origins of the Gradient,’’ American SHARELIFE Release Guide 1 (2010). http://www.share-project.org/share- Economic Review 92:5 (2002), 1308–1334. life/. Currie, Janet, ‘‘Healthy, Wealthy, and Wise: Socioeconomic Status, Poor Smith, James P., ‘‘Nature and Causes of Male Diabetes Trends, Undiag- Health in Childhood, and Human Capital Development,’’ Journal nosed Diabetes, and the SES Health Gradient,’’ PNAS 104:33 of Economic Literature 47:1 (2009), 87–122. (2007), 13,225–13,231. Deaton, Angus, ‘‘Height, Health, and Development,’’ PNAS 104:33 ——— ‘‘The Impact of Childhood Health on Adult Labor Market Out- (2007), 13232–13237. comes,’’ this review 91:3 (2009a), 478–489. Ellis, John, World War II: A Statistical Survey (London: Aurum Press, ——— ‘‘Re-Constructing Childhood Health Histories,’’ Demography 1994). 46:2 (2009b), 387–403. Gimbel, John, The American Occupation of Germany: Politics and the Statistisches Jahrbuch fur€ das Deutsche Reich (Statistical Yearbook), Military, 1945–1949 (Palo Alto, CA: Stanford University Press, Kaiserliches Statistisches Amt (Berlin: Puttkammer & Muhlbrecht,€ 1968). 1909–1939). Harrison, Mark, ‘‘The Economics of World War II: An Overview (PP. 1– Statistisches Bundesamt (1952, 1953), Statistisches Jahrbuch fur€ die Bun- 42), in Mark Harrison (ed.),’’ The Economics of World War II: Six desrepublik Deutschland (Stuttgart: W. Kohlhammer). Great Powers in International Comparison (Cambridge Cam- Van den Berg, Gerard, Pia A. Pinger, and Johannes Schoch, ‘‘Instru- bridge University Press, 1998). mental Variable Estimation of the Causal Effect of Hunger Early Heckman, James, ‘‘The Developmental Origins of Health,’’ Health Eco- in Life on Health Later in Life,’’ IZA discussion paper 6110 nomics 21:1 (2012), 24–29. (2011). Hionidou, Violetta, Famine and Death in Occupied Greece, 1941–1944 Van Mourik, W., Bilanz des Krieges (Rotterdam: Lekturama, 1978). (Cambridge: Cambridge University Press, 2006). Waldinger, Fabian, ‘‘Bombs, Brains, and Science. The Role of Human Ichino, Andrea, and Rudolf Winter-Ebmer, ‘‘The Long-Run Educational and Physical Capital for the Creation of Scientiﬁc Knowledge,’’ Cost of World War II,’’ Journal of Labor Economics 22:1 (2004), University of Warwick (2012). 57–87. Wehler, Hans-Ulrich, Deutsche Gesellschaftsgeschichte, 1914–1949 Jurges,€ Hendrik, ‘‘Collateral Damage: Educational Attainment and Labor (Munich: C. H. Beck, 2008). Market Outcomes among German War and Post-War Cohorts,’’ Zink, Harold, The United States in Germany, 1944–1955 (Princeton, NJ: unpublished manuscript (2011). Van Nostrand, 1957).

TABLE APPENDIX

TABLE A1.—EXTERNAL DATA VARIABLE DEFINITIONS Variable Name Deﬁnition Sources Casualties Total deaths Total casualties of World War II The data have been collected from four sources: Van Mourik (1978), Putzger (1963), Overman (1999), and Statistical Yearbook for the German Reich (1939) Total deaths/1939 Population Percentage of casualties in population Military deaths Military deaths in World War II Military deaths/1939 Population Percentage of military deaths in population Civilian deaths/1939 Population Civilian deaths of World War II Civilian deaths Fraction civilian deaths/ total population Holocaust deaths Holocaust deaths of World War II Holocaust deaths/1939 Population Fraction Holocaust deaths/total population

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/REST_a_00353 by guest on 29 September 2021 118 THE REVIEW OF ECONOMICS AND STATISTICS

TABLE A1.—(CONTINUED) Variable Name Deﬁnition Sources Population movement Source country Country from which people left Kulischer (1948) Target country Country to which people moved Time period Year in which people moved to target countries Number of combat operations War combat 0–2 months Respondent was living in a war country during the war period in a Ellis (1994) region within the country that experienced 0–2 months of combat War combat 3 or more months Respondent was living in a war country during the war period in a region within the country that experienced 3–10 months of combat Gross domestic product Log(GDP) Country-speciﬁc GDP Maddison (2011) Sex ratio Sex ratio Ratio of men to women for time periods before and after war Statistical Yearbooks for Germany (chaps. ‘‘Bewegung und Bevo¨lkerung’’, and ‘‘Internationale U¨ bersichten,’’ 1909–1939 1945–1953)

TABLE A2.—DESCRIPTIVE STATISTICS Percentage Missing All (N ¼ 21,266) War ¼ 0(N ¼ 11,411) War ¼ 1(N ¼ 9,855) Observations Mean SD Mean SD Mean SD Background information Year of birth 0.000 1942 9.52 1947 8.07 1936 7.13 Male 0.000 0.45 0.50 0.45 0.50 0.46 0.50 Childhood SES 0.034 0.02 1.00 0.22 0.99 0.21 0.96 Outcome measures Depression 0.000 0.36 0.48 0.34 0.47 0.38 0.48 Diabetes 0.002 0.10 0.30 0.07 0.25 0.13 0.34 Ever married 0.000 0.94 0.23 0.93 0.25 0.95 0.22 Heart disease 0.002 0.12 0.32 0.08 0.27 0.16 0.37 Height 0.007 168.31 8.85 169.47 8.92 166.97 8.59 Life satisfaction 0.081 7.60 1.72 7.87 1.62 7.29 1.79 Log(net worth) 0.040 12.45 1.87 12.94 1.83 11.88 1.75 Self-rated health 0.001 2.80 1.07 3.03 1.07 2.53 1.01 Years of education 0.080 10.72 4.16 11.71 3.88 9.56 4.18 Education difference in years 0.466 0.42 2.91 0.51 2.78 0.34 3.03 Channels of war exposure Dad absent 0.017 0.09 0.28 0.07 0.25 0.10 0.31 Dispossession 0.001 0.03 0.18 0.02 0.15 0.04 0.21 Hunger 0.001 0.07 0.25 0.02 0.14 0.13 0.33 Persecution 0.001 0.04 0.19 0.03 0.18 0.04 0.21 Childhood Immunizations 0.012 0.95 0.23 0.98 0.15 0.91 0.29 See table 1 for definitions of variables in first column and table A1 for a definition of war combat variables.

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/REST_a_00353 by guest on 29 September 2021